The Facts Capsule Template creates a simple capsule that stores lists of facts, optionally tagged to indicate their subject (that is, facts about world countries, facts about dried fruit, and so on). Users can then use this capsule to ask Bixby for a random fact, either from the entire list or with a specific tag.
Select the template in Bixby Developer Studio. Choose the language for your capsule. The next screen allows you to enter your capsule's facts.
playground.fruitFacts
.The capsule will be created for you in Bixby Studio.
This capsule is very simple: all the facts are stored in the content.js
file in the code
directory. The GetContent
action model returns a Content
structure. This Content
model corresponds to an item from the content.js
JSON file. The GetContent
action is tied to the GetContent.js
file through the endpoints.bxb
file using a local-endpoint
and returns a randomized item from content.js
.
For the example above, contents.js
would have a format like this:
module.exports = [
{
'text': 'Raisins are the most popular form of dried fruit.',
'tags': ['fruit'],
},
]
If the user request asks for a specific type of fact (for example, "fruit facts"), Bixby processes the optional SearchTerm
input. On the JavaScript side, the lib/util.js
filters the content with that search term through the tag
property.
For output, Bixby says one of two dialog options: either the text
from the returned content or the dialog that says "Nothing Found"
if no content is available. The capsule also provides a simple view.
Finally, the generated capsule contains one or two training entries to start with:
tag
", with tag
tagged as a SearchTerm
valueThe second entry will use your first tag as its example. If you didn't set any tags, this entry won't be created.
You can edit the content.js
file information with your own data, and add more content
items as needed. Just make sure that you follow the provided format.
If you are not linking to images hosted externally, then the local images must be in your assets/images
folder. You can sort these images into further subfolders, if necessary. For example, if you want to place all your animal images in one folder and all your plant images into another, you can place them under assets/images/animals
and assets/images/plants
respectively. You would then refer to each image URL as something like images/plants/tulip.png
.
If your content can be accessed through an API instead, you could choose not to use the content.js
file and instead Call a Web Service to return a JSON object that contains your material. However, you might have to change the existing modeling action implementations to process this and you might be better off creating a new capsule.
The template creates only the training entries given above. In order for Bixby to understand and properly reply to user requests for your capsule, you need to add additional training examples. You should pay particular attention to adding a few examples with tags that are related to your content
.
This section discusses some additional enhancements you can make that are not necessary for your capsule to be functional, but are optional enhancements you can do to further customize your capsule.
We provide dialog for content not found, but you can customize it further for your capsule. For example, if you're creating an animal facts capsule and a user requests an animal you don't have data for, you could have Bixby say I'm not koala-fied for this! or I'd be lion-ing if I had something for this!. You could also choose to randomize which dialog appears if this happens.
Make sure the dialog you write matches our Writing Dialog Design Guide.
The View file we currently supply has a simple image-list-of
component to help display any image associated with the content. However, you could further customize this. For example, if you are presenting plant facts, you could create an image card to show the type of plant you're telling a fact about as well as include the fact as part of the card's title-area
.
You could decide to add some input views. For example, if a user asks "Tell me a fact", Bixby could ask A fact about what? and present the options of bees, hot sauces, or fruit. The complexity of your user flow is up to you.
For more information on designing and updating the view, as well as best practices for user experience, you should read the Designing Conversations guides.